DPA-4.0.1-Pro-MPtrj

Version: v2026.06.05 Added: 2026-06-11 Published: 2026-06-11 22.8M parameters
Repo Paper Docs DOI Files PR Checkpoint
Predictions
DiscoveryGeometry OptimizationMolecular DynamicsκSRME
Leaderboard ranks CPS #15 /40Discovery F1 #23 /50Geo Opt RMSD #16 /41Phonons κSRME #14 /40MD CMDS #28 /34

Discovery: energy and convex hull diagnostics

Missing preds: 0

Per-element convex hull distance errors

1 H 0.18
2 He Helium
3 Li 0.03
4 Be 0.06
5 B 0.08
6 C 0.06
7 N 0.09
8 O 0.12
9 F 0.11
10 Ne Neon
11 Na 0.04
12 Mg 0.05
13 Al 0.07
14 Si 0.08
15 P 0.07
16 S 0.09
17 Cl 0.10
18 Ar Argon
19 K 0.04
20 Ca 0.05
21 Sc 0.04
22 Ti 0.05
23 V 0.07
24 Cr 0.10
25 Mn 0.12
26 Fe 0.11
27 Co 0.06
28 Ni 0.06
29 Cu 0.05
30 Zn 0.05
31 Ga 0.06
32 Ge 0.07
33 As 0.06
34 Se 0.10
35 Br 0.09
36 Kr Krypton
37 Rb 0.05
38 Sr 0.04
39 Y 0.05
40 Zr 0.05
41 Nb 0.06
42 Mo 0.06
43 Tc 0.05
44 Ru 0.09
45 Rh 0.07
46 Pd 0.07
47 Ag 0.04
48 Cd 0.04
49 In 0.07
50 Sn 0.06
51 Sb 0.06
52 Te 0.12
53 I 0.08
54 Xe 0.06
55 Cs 0.05
56 Ba 0.04
57 La 0.04
58 Ce 0.05
59 Pr 0.04
60 Nd 0.04
61 Pm 0.04
62 Sm 0.04
63 Eu 0.07
64 Gd 0.05
65 Tb 0.04
66 Dy 0.05
67 Ho 0.04
68 Er 0.04
69 Tm 0.05
70 Yb 0.05
71 Lu 0.04
72 Hf 0.05
73 Ta 0.09
74 W 0.07
75 Re 0.06
76 Os 0.08
77 Ir 0.09
78 Pt 0.07
79 Au 0.08
80 Hg 0.04
81 Tl 0.05
82 Pb 0.07
83 Bi 0.06
84 Po Polonium
85 At Astatine
86 Rn Radon
87 Fr Francium
88 Ra Radium
89 Ac 0.04
90 Th 0.06
91 Pa 0.09
92 U 0.07
93 Np 0.14
94 Pu 0.32
95 Am Americium
96 Cm Curium
97 Bk Berkelium
98 Cf Californium
99 Es Einsteinium
100 Fm Fermium
101 Md Mendelevium
102 No Nobelium
103 Lr Lawrencium
104 Rf Rutherfordium
105 Db Dubnium
106 Sg Seaborgium
107 Bh Bohrium
108 Hs Hassium
109 Mt Meitnerium
110 Ds Darmstadtium
111 Rg Roentgenium
112 Cn Copernicum
113 Nh Nihonium
114 Fl Flerovium
115 Mc Moscovium
116 Lv Livermorium
117 Ts Tennessine
118 Og Oganesson
57-71 La-Lu Lanthanides
89-103 Ac-Lr Actinides

ML vs DFT Lattice Thermal Conductivity

Model Authors

  1. Tiancheng Li  AI for Science Institute, Beijing; Peking University  
  2. Duo Zhang  AI for Science Institute, Beijing  
  3. Linfeng Zhang  AI for Science Institute, Beijing; DP Technology  
  4. Han Wang  Beijing Institute of Applied Physics and Computational Mathematics (IAPCM)  

Trained By

  1. Tiancheng Li  AI for Science Institute, Beijing; Peking University  

Model Info

  • Version v2026.06.05
  • Role Interatomic potential
  • Architecture gnn
  • Targets EFSG
  • Openness OSOD
  • Discovery Train Task S2EFS
  • Discovery Test Task IS2RE-SR

Training Set

MPtrj: 1.58M structures from 146k materials

description

DPA-4.0.1-Pro-MPtrj is the DPA4-Pro universal interatomic potential trained only on the MPtrj dataset for this Matbench Discovery submission.

architecture

DPA4 is an SE(3)-equivariant potential built on an EMFA (Edge-conditioned, Multi-Focus, Attention) SO(2)-equivariant convolution: a low-rank edge-node SO(2)-equivariant product, a multi-focus design for message nonlinearity, and envelope-gated attention for message aggregation, with a Lebedev-grid projection that preserves SO(3)-equivariance in the nonlinearity.

training

The model was trained for 2,000,000 steps on 16 GPUs using the HybridMuon optimizer, WSD learning-rate schedule, MAE energy/force/virial loss weights 20/20/5, with bf16 AMP, TF32 matmul, and torch compile enabled.

Hyperparams

  • evaluation: {"max_force":0.02,"max_steps":500,"ase_optimizer":"FIRE","cell_filter":"FrechetCellFilter","kappa":{"protocol":"phonondb-v1","displacement_distance":0.03,"save_forces":true}}
  • architecture: {"graph_construction_radius":6,"max_neighbors":384}
  • upstream_config: {"architecture":{"type":"DPA4/SeZM","feature_dim":64,"n_focuses":2,"n_layers":6,"so2_layers":4,"ffn_layers":2,"radial_basis":"Bessel","n_radial_basis":16,"lmax":5,"mmax":1,"edge_node_product":"degree mixing","per_channel_modulation":true,"rank":2,"attention_heads":1,"s2_activation":"FFN only","quadrature":"Lebedev","norm_placement":"Post & Pre","activation_function":"SiLU","ffn_hidden_dim":"auto","output_fitting_dim":"auto","output_fitting_layers":1,"precision":"float32"}}
  • training: {"optimizer":"HybridMuon","muon_mode":"slice","magma_lite":true,"weight_decay":0.001,"lr_scheduler":"WSD","max_lr":0.0005,"min_lr":0.000001,"warmup_steps":5000,"decay_ratio":0.65,"decay_type":"cosine","loss":"MAE","loss_weights":{"energy":20,"force":20,"virial":5},"batch_size_per_gpu":"filter:300","training_steps":2000000,"gradient_max_norm":5,"n_gpus":16,"compile":true,"bf16_amp":true,"tf32_matmul":true}

Dependencies